16 papers

Foundations & Infrastructure

The substrate: LLM internals, GPUs, serving systems, and the theory underneath.

Papers
16
Read time
5.3h
Avg length
20m
01
Foundations & Infrastructure

Fundamentals of Building Autonomous LLM Agents

A practitioner's anatomy of an autonomous LLM agent — it decomposes the agent into four cognitive subsystems (perception, reasoning/planning, memory, execution), catalogs the concrete techniques for each, and shows how to wire them into a closed loop that can actually drive a GUI or web task end-to-end.

VICTOR DE LAMO CASTRILLO, HABTOM KAHSAY GIDEY, ALEXANDER LENZ, ALOIS KNOLL (UPC BARCELONA / TU MÜNCHEN) · OCT 2025
Read
16 min
Source
arXiv:2510.09244
02
Foundations & Infrastructure

Building Autonomous AI Agents based AI Infrastructure

Put autonomous agents in charge of the boring, expensive, 3-AM parts of running cloud/AI infrastructure — scaling, fault recovery, energy, security — using a perceive → decide → act → learn loop, and a single RL-driven agent beats both rule-based automation and humans on utilization, recovery time, and energy.

APURVA KUMAR (WALMART GLOBAL TECH) · IJCTT, VOL 72 ISSUE 11, 2024 · DOI:10.14445/22312803/IJCTT-V72I11P112
Read
14 min
03
Foundations & Infrastructure

@skills: Attention Is All You Have

Installing an agent skill secretly bundles three separate jobs — fetch the text, keep the text, and let it fire on its own — but only the last one needs a permanent slot in the model's system prompt, so `@skills` splits them apart and lets 56,804 skills be reachable while fewer than ten sit resident.

LI YIN, ZHI LI, ZHAN SHI, HAORAN ZHANG, HAEBIN SEONG, ZHANGYANG WANG (SYLPHAI · UT AUSTIN)
Read
18 min
Source
arXiv:2608.12610
04
Foundations & Infrastructure

Adaptive LLM Routing under Budget Constraints

A query router that learns *online* — from cheap thumbs-up/down feedback instead of a labeled dataset — which LLM to send each query to, warm-started with human-preference data and capped by a budget-aware "knapsack" policy, hitting 93% of GPT-4's quality at 25% of its cost.

PANDA, MAGAZINE, DEVAGUPTAPU, TAKEMORI & SHARMA (FUJITSU RESEARCH / MICROSOFT)
Read
16 min
Source
arXiv:2508.21141
05
Foundations & Infrastructure

Agentic Web: Weaving the Next Web with AI Agents

The web is mutating from a place humans search and click into a place where autonomous agents discover each other, call services, and transact on your behalf — and this survey maps the three layers (intelligence, interaction, economics) and the new plumbing (MCP, A2A, agent-attention markets) you'll need to build in that world.

YANG, MA, HUANG, ET AL. (SJTU, HKUST, BERKELEY, UCL, ET AL.)
Read
28 min
Source
arXiv:2507.21206
06
Foundations & Infrastructure

Efficient Agents: Building Effective Agents While Reducing Cost

A controlled ablation of every part of an LLM agent (backbone, planning, tools, memory, test-time scaling) on the GAIA benchmark, showing that most of the "sophisticated" machinery adds cost without accuracy — and that a deliberately stripped-down config keeps ~97% of performance for ~57% of the cost.

OPPO AI AGENT TEAM · 2025
Read
14 min
Source
arXiv:2508.02694
07
Foundations & Infrastructure

Fast and Simplex: 2-Simplicial Attention in Triton

Replace the pairwise dot-product inside attention with a three-way (trilinear) product so each model parameter buys more reasoning power — letting you hit better math/code/logic scores with the same number of training tokens, which matters now that high-quality data, not compute, is the bottleneck.

AURKO ROY, SIJIA CHEN, TIMOTHY CHOU, JIECAO YU, SAI SURYA DUVVURI, XIAODONG WANG, ROHAN ANIL, MANZIL ZAHEER (META) · 2025
Read
16 min
Source
arXiv:2507.02754
08
Foundations & Infrastructure

Foundations of Large Language Models

A 277-page textbook that walks you through the entire lifecycle of a modern LLM — pre-train it, scale it, prompt it, align it to humans, and squeeze more reasoning out of it at inference time — explaining the actual math behind each stage rather than just naming the techniques.

TONG XIAO & JINGBO ZHU (NLP LAB, NORTHEASTERN UNIVERSITY & NIUTRANS RESEARCH) · 2025
Read
45 min
Source
arXiv:2501.09223
09
Foundations & Infrastructure

From LLM Inference to Agentic Workloads

When you run real AI agents at scale, the language model is no longer the main cost — the sandboxes, search, retrieval, idle-but-live sessions, and framework overhead are — and this paper measures exactly where the time, memory, and money actually go, then shows four small fixes that reclaim 29–460% of it.

CHANG, ZHOU, FU, ET AL. (HKUST, ALIBABA, BYTEDANCE)
Read
18 min
Source
arXiv:2608.15127
10
Foundations & Infrastructure

GPUs: How They Work and How to Make Them Fast

A modern GPU is a factory that can compute far faster than it can fetch data, so almost every trick for making machine-learning code fast is really a trick for touching slow memory less often — and FlashAttention is the poster child of doing exactly that.

TATSU HASHIMOTO · CS336 (LANGUAGE MODELING FROM SCRATCH), STANFORD · LECTURE 5
Read
22 min
11
Foundations & Infrastructure

Language Models are Injective and Hence Invertible

A decoder-only Transformer almost never maps two different prompts to the same hidden state, so the hidden states you store, log, or cache *are* the user's text in disguise — and the paper gives you a linear-time algorithm (SipIt) that reads it straight back out.

NIKOLAOU, MENCATTINI, CRISOSTOMI, SANTILLI, PANAGAKIS, RODOLÀ (SAPIENZA / EPFL / ATHENS / ARCHIMEDES) · 2025
Read
18 min
Source
arXiv:2510.15511
12
Foundations & Infrastructure

Persona Vectors: Monitoring and Controlling Character Traits in Language Models

Find the single direction in a model's activations that "means" a trait like *evil* or *sycophancy*, and you get a cheap dial to monitor it live, undo it after fine-tuning, prevent it during fine-tuning, and pre-screen training data that would induce it — all from a one-sentence description of the trait.

RUNJIN CHEN, ANDY ARDITI, HENRY SLEIGHT, OWAIN EVANS, JACK LINDSEY (ANTHROPIC FELLOWS / UT AUSTIN / UC BERKELEY / TRUTHFUL AI / ANTHROPIC) · JUL 2025
Read
16 min
Source
arXiv:2507.21509
13
Foundations & Infrastructure

Ray: A Distributed Framework for Emerging AI Applications

Ray is a distributed compute framework that runs millions of tiny, heterogeneous tasks per second with millisecond latency by combining a stateless "remote function" model, a stateful "actor" model, a logically-centralized control store, and a two-tier scheduler — making it the substrate that lets you scale RL and agentic workloads with a few `@ray.remote` decorators instead of a custom cluster.

MORITZ, NISHIHARA, WANG, TUMANOV, LIAW, LIANG, PAUL, JORDAN, STOICA (UC BERKELEY) · ARXIV 2017 · 1712.05889
Read
18 min
14
Foundations & Infrastructure

Role-Aware Language Models for Secure and Contextualized Access Control in Organizations

Can you bake org-chart permissions directly into an LLM so it answers the CEO but politely refuses the intern asking the same question? This paper fine-tunes models to do exactly that — and measures how well it survives jailbreaks, prompt injection, and corrupted role labels.

ALMHEIRI, KONGRAT, SANTOSH, TASMUKHANOV, LOZA VERA, AL KAUTSAR, KOTO (MBZUAI ET AL.) · 2025
Read
16 min
Source
arXiv:2507.23465
15
Foundations & Infrastructure

Towards a Science of Scaling Agent Systems

A controlled 180-run study showing that adding agents helps or hurts in predictable ways governed by measurable task properties — and distilling that into one equation that picks the right architecture for 87% of unseen tasks.

KIM, GU, PARK ET AL. (GOOGLE RESEARCH, GOOGLE DEEPMIND, MIT) · DEC 2025
Read
18 min
Source
arXiv:2512.08296
16
Foundations & Infrastructure

Understanding Deep Learning

A single, internally-consistent textbook that teaches every modern deep-learning architecture — MLPs, CNNs, transformers, GNNs, GANs, VAEs, diffusion, and RL — as variations on one recipe (define a parameterized function, write a loss, minimize it by gradient descent), so you finish able to reason about *why* these systems work, not just call them.

SIMON J.D. PRINCE · MIT PRESS, 2023 (UDLBOOK.COM)
Read
28 min